Enhancing the Situation Awareness of Decision Makers by Applying Case-Based Reasoning on Streaming Data

Abstract

Data is generally expected to continue its exponential growth the next five to ten years. However, it is commonly agreed that the amount of data that currently exist is abundant and that there is still much to achieve with it. The industry, in general, has recognized both the growth and the need to analyze the data. This is also the case with oil well drilling operations. Currently, operators monitor oil well drilling operations manually by staring at real-time measurements visualized in a computer program as graphs. Having humans monitoring the drilling operations manually has its disadvantages, as people get bored, tired and distracted. In this thesis, we investigate whether real-time decision making can be improved by enhancing the decision maker’s situation awareness through applying case-based reasoning on streaming data. This thesis is not about automating decisions, but informing human decision makers about the current situation so that they can make an informed decision. A hybrid reasoning system that abstracts recognize symptoms in time-series data and describe the current situation using these is described. The current situation is compared to past problematic situations and the similar past situations are brought to the attention of the decision maker to support decisions. Furthermore, situation assessment, the process of acquiring an understanding of the current state of a situation, in oil well drilling is analyzed and described. There are four main contributions of the research effort presented in this thesis: (i) a case representation for drilling situations; (ii) a hybrid reasoning architecture that is capable of reasoning with real-time data streams; (iii) a similarity metric for sequences of complex events; and (iv) a knowledge level model of situations and situation assessment.